AI could consume nearly half of global datacentre power by year-end, new analysis warns

Artificial intelligence systems can represent approximately half of all energy consumption in global data centers by the end of this year, according to the new research – providing increasing concerns about the environmental impact of artificial intelligence technologies.

The analysis, conducted by Alex de Fais Gao, the founder of the classification technology platform, notes that artificial intelligence can represent up to 49 % of the total use of data energy at the end of 2025. The study is scheduled to be published in Energy Journal and comes after days of international expectations (IEA) that AI can require almost through the end of December.

Based on the electricity drawn by chips from the main AI service providers including NVIDIA, AMD and Broadcom, research is estimated that artificial intelligence currently represents about 20 % of the total energy consumption in data centers – an important segment of 415 hours of Tira loo Subscribers.

De vries-gao was taken into account in variables such as devices efficiency, cooling systems and the density of work burden to estimate the increasing share of the demand. It warns that the frequency of expansion of AI and the model can soon push AI’s energy consumption to 23 GB-more than twice the total energy use in the Netherlands.

“These innovations can reduce the costs of arithmetic and energy for Amnesty International,” said De Fais Gao. “But efficiency gains can also encourage broader dependence – and ultimately more energy -used.”

The analysis comes amid a rapid increase in sovereign artificial intelligence initiatives, as the two countries invest in their infrastructure from artificial intelligence – a trend that is likely to increase the demand for global devices. One of the examples of cited is Crusoe Energy, a US -based startup that has recently received 4.5 GB of gas capacity in new databases, with a possible customer through the joint Stargate project.

On Thursday, Openai confirmed the launch of the first Stargate facility outside the United States, in the United Arab Emirates. De vries-gao warned that such developments can exacerbate the dependence on fossil fuels, undermining the green aspirations to lead the artificial intelligence companies.

Microsoft and Google recognized that aggressive artificial intelligence efforts threaten their internal environmental goals, as the energy imprint grows in the burdens of artificial intelligence action beyond expectations.

Despite the increasing concerns, de Fis Gao said that the data related to the consumption of operational energy from artificial intelligence is still rare, describing the Strip as a “transparent industry.” Although the European Union Law of Amnesty International will soon require companies to reveal the consumption of training power, it does not impose reports on the energy used to operate the artificial intelligence models daily – a major shareholder in continuous emissions.

“We urgently need more transparency about the energy cost of Amnesty International,” said Professor Adam Soby, Director of Sustainability at the Alan Torring Institute, The UK Institute for Intelligence Research.

Sobby added that despite the high energy consumption in the front end of the artificial intelligence, technology still plays a role in reducing carbon emissions in other places, especially in sectors such as transportation and energy, where behavior improvement tools that work can lead to great savings.

“I think we do not need many very good use cases to compensate for the energy used in the front end,” Sobi said.

While governments, investors, and companies pressure the development of artificial intelligence, the results emphasize the need for more vision, organization and innovation to achieve a balance between artificial transformational intelligence promise with the fingerprint of the growing environment.